Note: This discussion is about an older version of the COMSOL Multiphysics® software. The information provided may be out of date.

Discussion Closed This discussion was created more than 6 months ago and has been closed. To start a new discussion with a link back to this one, click here.

turbulent flow kappa-epsilon, 3D

Please login with a confirmed email address before reporting spam

Hi

I need to model high Reynolds flow in 3D. The model is very simple: it consists of one quarter of a cylinder (using symmetry). The density is 1.23 kg/m^3; viscosity: 1.79e-5; velocity at the inlet: 16 m/s; radius of the cylinder: 0.75 in; length: 7.5 in. For the inlet, turbulent intensity is 0.03 and turbulence length scale is 0.07*2*radius. The model is attached.

The model doesn't run. I get two warnings:
"Ill-conditioned preconditioner. Increase factor in error estimate to 43.311" and
"New constrain force nodes detected: these are not stored!"

I don't understand what these warnings mean, so I don't know where to start. I'm also new at turbulence and boundary layers, so I don't know if the problem would be there.

Any assistance is greatly appreciated, since the geometry I'm aiming for in the near future is much more complicated!

Thank you,
Sylvana



3 Replies Last Post 26 gen 2011, 10:48 GMT-5
Ivar KJELBERG COMSOL Multiphysics(r) fan, retired, former "Senior Expert" at CSEM SA (CH)

Please login with a confirmed email address before reporting spam

Posted: 1 decade ago 22 gen 2011, 15:35 GMT-5
Hi

I have had a rapid look at the model, I have a few comments:

My experience is that the default initial values "0" are often too far from the final solution so the solver looses it's way before managing to get to some reasonnable convergence slope.

There are a few things you can easily implement:

1) try not to use "r" and single letters for your own variable names, as most of these are already used by COMSOL, so I changed your "r" by "rin". Note: in your case you do pobably NOT have an issue, but it's by chance

2) define a cylindrical coordinate system for such symmetry problems then you have acces to sys2.r the radial direction (another reason to avoid "r" as user parameter

3) put some initial conditions different from all "0", at least use a parabolic speed distribution like:

vx=vy=0 and vz = Vinl*2*(1-(sys2.r/rin)^2)[m/s]

and for the pressure, use at least Poiseille head loss pressure distribution

p= 8*etaf*Vinl/rin^2*(10*rin)*(1-z/(10*rin))

in your notation where 10*rin is the total tube length and "etaf" the viscuosity of the fluid. These are normally for laminar flow, but are far better than all "0"

then you should upgrade your COSMOL version with th patch on the main COMSOL site to at least v4.1.0.112, this fixes a few internal formula issues.

Then I Deleted the solution, generated a default solver sequence, selected the "Dependent Variables" node right click "Solve to selected" and observed the velocity and the pressure of the initial values, to be sure they are correct. And I run the programme as is, it solved in about 10 min.
And your CellReynolds number seem reasonnable for me so the mesh is probably OK, but you could also try with the default phyiscal induced mesh.

But are you sure you are really "turbulent" here ?

--
Good luck
Ivar
Hi I have had a rapid look at the model, I have a few comments: My experience is that the default initial values "0" are often too far from the final solution so the solver looses it's way before managing to get to some reasonnable convergence slope. There are a few things you can easily implement: 1) try not to use "r" and single letters for your own variable names, as most of these are already used by COMSOL, so I changed your "r" by "rin". Note: in your case you do pobably NOT have an issue, but it's by chance 2) define a cylindrical coordinate system for such symmetry problems then you have acces to sys2.r the radial direction (another reason to avoid "r" as user parameter 3) put some initial conditions different from all "0", at least use a parabolic speed distribution like: vx=vy=0 and vz = Vinl*2*(1-(sys2.r/rin)^2)[m/s] and for the pressure, use at least Poiseille head loss pressure distribution p= 8*etaf*Vinl/rin^2*(10*rin)*(1-z/(10*rin)) in your notation where 10*rin is the total tube length and "etaf" the viscuosity of the fluid. These are normally for laminar flow, but are far better than all "0" then you should upgrade your COSMOL version with th patch on the main COMSOL site to at least v4.1.0.112, this fixes a few internal formula issues. Then I Deleted the solution, generated a default solver sequence, selected the "Dependent Variables" node right click "Solve to selected" and observed the velocity and the pressure of the initial values, to be sure they are correct. And I run the programme as is, it solved in about 10 min. And your CellReynolds number seem reasonnable for me so the mesh is probably OK, but you could also try with the default phyiscal induced mesh. But are you sure you are really "turbulent" here ? -- Good luck Ivar

Please login with a confirmed email address before reporting spam

Posted: 1 decade ago 26 gen 2011, 00:27 GMT-5
Hi, Ivar

Thank you for your advice. Just changing the r to rin and entering a value for initial values (I just tried the value for the inlet velocity), the model ran. However, I chnged the global parameters, still maintaining the Re constant (40000) and it didn't run. I changed back these parameters and it didn't work either. Although the progress bar advances and even finishes, there's no data set generated in the Model Builder. Before running another model, I go to "Solver configurations" and then "Clear solutions". Is there anything else I need to do?

Thank you,
Sylvana

P.S. I am working on the upgrade to 4.1.0.112.
Hi, Ivar Thank you for your advice. Just changing the r to rin and entering a value for initial values (I just tried the value for the inlet velocity), the model ran. However, I chnged the global parameters, still maintaining the Re constant (40000) and it didn't run. I changed back these parameters and it didn't work either. Although the progress bar advances and even finishes, there's no data set generated in the Model Builder. Before running another model, I go to "Solver configurations" and then "Clear solutions". Is there anything else I need to do? Thank you, Sylvana P.S. I am working on the upgrade to 4.1.0.112.

Ivar KJELBERG COMSOL Multiphysics(r) fan, retired, former "Senior Expert" at CSEM SA (CH)

Please login with a confirmed email address before reporting spam

Posted: 1 decade ago 26 gen 2011, 10:48 GMT-5
Hi

note there isa difference between CLEAR solution (delete data but keep process) and DELETE solution & Sequence which also delete the solver sequence, AND all results depending on this solver sequence.

If you want to keep your results sequence, define a new dummy study, generate sequence, get initial values (to full the matrices and generate a solution) link all results to the now solutions of this new study, then only delete the solver sequence, and regenerate it from fresh.

If you edit, particularly in v4, the solver sequence often it sometimes get corrupt, or some links do not follow. therefore I mostly "delete" my solver sequence, then often it solves again

--
Good luck
Ivar
Hi note there isa difference between CLEAR solution (delete data but keep process) and DELETE solution & Sequence which also delete the solver sequence, AND all results depending on this solver sequence. If you want to keep your results sequence, define a new dummy study, generate sequence, get initial values (to full the matrices and generate a solution) link all results to the now solutions of this new study, then only delete the solver sequence, and regenerate it from fresh. If you edit, particularly in v4, the solver sequence often it sometimes get corrupt, or some links do not follow. therefore I mostly "delete" my solver sequence, then often it solves again -- Good luck Ivar

Note that while COMSOL employees may participate in the discussion forum, COMSOL® software users who are on-subscription should submit their questions via the Support Center for a more comprehensive response from the Technical Support team.